Ask ten superintendents what "predictable" means and you'll get ten shrugs, because most of us have made peace with the idea that construction just doesn't finish when it's supposed to. You pad the schedule, you carry a mental buffer nobody sees, and when the job lands within a couple weeks of the promise you call it a win. That's not predictability. That's a controlled skid.
The Last Planner System exists because a few people got tired of the skid and asked a sharper question: if we can't reliably predict a whole project, can we at least reliably predict next week? Turns out you can. And once you can predict a week, you can stack predictable weeks into a predictable project. That's the whole game. The software just makes it possible to run the discipline without drowning in paperwork.
Why construction feels like a coin flip
Unpredictability isn't bad luck. It comes from a handful of habits we all recognize, because we've all done them.
We build schedules out of hope. The CPM baseline says drywall starts Monday because the logic says it should, not because anyone confirmed the framing inspection cleared, the material's on the deck, and the crew isn't stuck finishing another floor. We commit to work before we've checked whether the work is actually ready.
Then we discover the constraints during execution instead of before it — the RFI that's still open, the light fixtures that are eight weeks out, the sequence conflict where the ceiling grid and the overhead MEP both need the same airspace on the same day. Every one of those turns into a stoppage, and stoppages cascade. One trade slips two days, the next trade's crew is already booked on another job, and now you've lost a week off a two-day problem.
Underneath all of it is the quiet killer: unreliable commitments. A sub says "yeah, we'll be there Wednesday" in a meeting, and everyone writes it down like it's real. It isn't. It's a wish said out loud. When the plan is built on a stack of soft promises, of course the outcome is a coin flip.
The math nobody wants to look at
Here's the part that should make you sit up, and it's the one genuinely important idea worth keeping from any discussion of reliability.
Say each activity in your schedule has a 70% chance of going as planned. That's roughly what traditional field scheduling delivers — two out of three commitments hit, one whiffs. Sounds tolerable. But activities don't live alone; they live in chains. Drywall waits on framing, framing waits on rough-in, rough-in waits on the deck being clear. Chain ten of those 70% activities together and the odds that the whole chain runs clean is 0.7 raised to the tenth power — under 3%. Effectively zero. That's why the far end of your schedule is always fiction.
Now push each activity to 90% reliability. Same ten-activity chain: 0.9 to the tenth is about 35%. Still not a sure thing, but you've gone from "never" to "better than one in three," and every additional point of reliability compounds hard across the chain. This is the case for the Last Planner System in one line: the leverage isn't in the master schedule, it's in raising the reliability of individual weekly commitments. Get each promise from 70% to 90% and the whole project's behavior changes.
So how do you actually move that number?
Make-ready: kill the constraint before it kills the week
The single biggest reason a committed task fails is that it was never truly ready to start. Last Planner attacks this with the lookahead — a rolling three-to-six-week window where you screen upcoming work for constraints and knock them down before the work reaches the weekly plan.
The constraint categories are always the same handful, and you should be checking every upcoming task against them:
- Prior work — is the predecessor actually complete and inspected, not just "mostly done"?
- Design/information — any open RFI, missing detail, or unapproved submittal touching this scope?
- Materials — on site, or a confirmed delivery that lands with a day or two of slack, not the morning of?
- Labor — the right crew, the right size, actually available, not double-booked?
- Equipment — lift, hoist, pump, whatever the scope needs, reserved?
- Access & conditions — area released, safe, dried in, powered, at temperature for the work?
A task doesn't earn a spot on the weekly work plan until every one of those is clear. That's the rule that does the heavy lifting. The lookahead is where you find the eight-week light fixtures in week six instead of week zero. It's where you catch that two trades are stacked in the same room and you split the sequence before either one shows up. A rolling lookahead schedule in something like LookAheadWall makes that screening a routine you run every week instead of a heroic effort you do once and abandon — you drag work into the near-term window, and the constraints ride along visibly with it so nothing gets committed blind.
One practical habit: assign every open constraint an owner and a need-by date, and treat that list like a hit list. A constraint with no name attached to it doesn't get resolved; it gets rediscovered — on the day the work was supposed to start.
Commitments that mean something
The weekly work plan is where the last planners — the foremen and crew leaders who actually direct the work — make promises to each other for the coming week. The word "commitment" is doing real work here. A good commitment is:
- Made by the person doing the work, not assigned to them from above. If the foreman didn't promise it, it isn't a commitment, it's an order, and orders miss.
- Constraint-free — it already cleared make-ready, so nothing outside the crew's control should block it.
- Sized to real capacity — based on the crew you'll actually have Tuesday, not the crew you wish you had.
- Specific enough to score — "make progress on level 3" is unscoreable; "hang and fire-tape the north corridor, grids 4–9" is a promise you can check off or not.
Keep a small workable backlog too — ready, constraint-free tasks that aren't on the critical path. When the primary work stalls for a reason you couldn't foresee, you slide the crew onto backlog instead of sending them home. That's the difference between a bad hour and a lost day.
PPC: the number that tells you the truth
At the end of the week you count how many committed tasks were fully done, divide by how many you committed to, and that's your Percent Plan Complete. Committed to twenty, finished seventeen, that's 85%. It's a blunt instrument — half a task counts as zero, no partial credit — and that bluntness is the point. It won't let you lie to yourself.
PPC is your predictability number. High PPC means your weekly plan is a reliable forecast of what will actually happen, which is the entire definition of predictable. Most crews start in the 50s when they first measure honestly, and that's fine — you can't fix what you won't count. Good teams climb into the 80s and hold. If you're sitting at a genuine, sustained 85%+, your commitments are running near that 90% reliability the math wants, and the far end of your schedule stops being fiction.
But the PPC number alone is worth little. The gold is in the misses. Every unfinished commitment gets a reason code — prerequisite not done, material late, weather, crew reassigned, design conflict, plain over-commitment. Tally those reasons over a month and a pattern jumps out. Maybe 40% of your misses trace back to one sub's prerequisites running late every single week. Now you're not managing a vague feeling that the job's behind; you're solving one specific, repeating root cause. That variance analysis is where the reliability actually gets built, week over week. Tracking PPC and reason codes by hand in a spreadsheet is doable but painful; letting the scheduling software roll it up automatically is the difference between a team that reviews the data and one that means to.
How weekly reliability becomes a delivery date you can stand behind
Reliable weeks compound. Once you've got a few months of honest PPC history, you know your crew's real throughput — not the estimator's number, the measured one. That lets you forecast the milestones ahead with something better than optimism, and it lets you connect the six-week horizon back to the master schedule with evidence instead of hope.
This is what makes the practice worth the trouble to everyone standing around the schedule:
- The owner gets an occupancy date they can plan operations and financing around, because it's built on measured reliability, not a bar chart's wishful geometry.
- The GC gets fewer fire drills — crews and equipment show up when the work is genuinely ready, so you stop paying men to stand in a room they can't work in.
- The subs get a schedule they can actually staff to. A sub who's been burned by three false starts learns to pad and hedge; a sub who trusts your dates commits their best crew. Reliable planning is how you earn that trust, and it pays you back in manpower.
What quietly destroys it
The system fails in predictable ways, and they're almost always cultural before they're technical:
- Optimism pressure. Someone up the chain leans on the plan until it shows the date they want instead of the date the work supports. Now the plan lies, and everyone knows it, and the whole exercise rots.
- Skipping make-ready. When the lookahead gets rushed or dropped, constraints stop getting caught early and you're back to discovering them on the deck.
- Over-committing. Loading the weekly plan with everything you'd like done guarantees a low PPC and teaches the crew the plan is theater.
- Ignoring the reason codes. Measuring PPC and never acting on the misses is just paperwork. The number without the follow-up changes nothing.
The bottom line
Predictability isn't a personality trait some jobs are born with and others aren't. It's the arithmetic of a lot of small, reliable commitments stacked on top of each other. Screen the work in the lookahead so it's genuinely ready. Let the people doing the work make the promises. Measure whether the promises held, and go fix the reasons they didn't. Do that consistently and the number climbs, and as the number climbs, the job starts finishing when you said it would.
Software doesn't create that discipline — no tool ever has. What good look-ahead scheduling does is take the friction out of running it every single week, so the make-ready screening, the weekly commitments, and the PPC tracking actually happen instead of being the thing you'll start doing next month. Get the discipline right and the predictability isn't luck. It's just what happens.